Investigation of the Trap States in Fully-Recessed Normally off LPCVD-Si<inf>3</inf>N<inf>4</inf>/PEALD-AlN/GaN MIS-HEMT with in-Situ N<inf>2</inf> or H<inf>2</inf>/N<inf>2</inf> Plasma Pretreatment

Jiaofen Yang,Ming Tao,Jing Xiao,Bin Zhang,Hongyue Wang,Min He,Jie Liu,Jinyan Wang,Maojun Wang
DOI: https://doi.org/10.1109/SSLChinaIFWS60785.2023.10399730
2023-01-01
Abstract:Fully recessed normally off LPCVD-Si3N4/PEALD-AlN/GaN MIS-HEMTs are fabricated with self-terminating wet etch technique. The effects of in-situ N2 or H2/N2 plasma treatments prior to the deposition of PEALD-AlN on the trap states of the device are studied. The interface trap density(Dit) is extracted using the multi-frequency capacitance-voltage (MFCV) method, whereas the border trap density (Nt) is extracted using the low frequency noise (1/f noise) method. The total Dit of devices with in-situ N2 and H2/N2 plasma pretreatment are 9.86 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">12</sup> and 8.51 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">12</sup> eV <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> , respectively, an order of magnitude lower than no plasma pretreatment (1.52 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">13</sup> eV <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-2</sup> ). The Nt of devices with no plasma pretreatment, in-situ N2 and H2/N2 plasma pretreatment are 6.20 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">19</sup> ∼ 7.26 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">19</sup> , 5.04 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup> ∼ 7.97 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup> and 3.05 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup> ∼ 4.06 × 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup> eV <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-3</sup> , respectively. These results indicate that the in-situ H2/N2 plasma pretreatment is more effective at reducing the trap density compared to in-situ N2 plasma pretreatment. Besides, it is found that the plasma pretreatment process may suppress border trap better than interface trap.
What problem does this paper attempt to address?